Video to SRT: Generate + Add Subtitles to Any Video (2026 Guide)

Video to SRT at a Glance

The phrase “video to SRT” hides two very different tasks, and users searching for it split roughly evenly between them. The first job is generation: you have a raw video file and need a subtitle file that matches the spoken content, cue by cue, with per-line timestamps in HH:MM:SS,mmm format. The second job is combination: you already have both files — a finished SRT and a finished video — and you want them to travel together, either as separate soft-sub tracks that viewers toggle or as hardcoded captions rendered permanently into the video pixels. Different tools solve each. This page walks both honestly, with real accuracy numbers, format compatibility, decision matrices, and FFmpeg commands you can copy directly.

Which Job Do You Have?

Pick the workflow that matches the files you actually have on disk right now.

Job A: Generate SRT from Video

Five-step workflow to convert a video file into a valid SRT subtitle file with per-cue timestamps. Total wall-clock time for a 1-hour video: 10-15 minutes end-to-end, most of which is Whisper Large-v3 processing.

Job B: Add an Existing SRT to Video

You have a finished SRT and a finished video, both on disk. The next decision is a fork: soft-subs or burn-in. The two paths solve different problems and have different trade-offs. The decision matrix in §E lays them out in a table; this section walks each path operationally.

ffmpeg -i input.mp4 -i input.srt -c copy -c:s mov_text output.mp4

ffmpeg -i input.mp4 -vf "subtitles=input.srt" output.mp4

Burn-in vs Soft-subs: Decision Matrix

Head-to-head across the seven dimensions that actually decide the fork. If your platform supports soft-subs and your content is long-form, soft-subs win. If your platform does not (TikTok, Instagram Reels), or the captions are load-bearing for feed engagement, burn-in wins.

Source: WCAG 2.1 SC 1.2.2 (June 2018) accepts either; verified July 2026.

Video Format Compatibility

Whisper Large-v3 does not care about the video container — it operates on the audio track. VexaScribe extracts audio via FFmpeg before invoking Whisper, so anything FFmpeg can decode works. In practice, that covers every container format used in the modern web-video pipeline.

Source: Whisper Large-v3 accepts any format FFmpeg can decode; verified July 2026.

MP4 (H.264 + AAC) is our recommended input because it is universally decodable, well-optimized in modern FFmpeg builds, and matches what almost every camera and screen recorder outputs. MOV files from iPhone (typically HEVC + AAC) also work but occasionally have variable-frame-rate quirks that older FFmpeg builds handle poorly; a modern FFmpeg (7.x, released 2024) resolves them. If you have a container we do not list — OGV, FLV, 3GP — it almost certainly still works; FFmpeg's demuxer list is long.

Accuracy by Video Scenario

Video-to-SRT accuracy is fundamentally audio accuracy — Whisper Large-v3 does not look at the picture. That means the honest breakdown depends on what kind of audio your video has: single-speaker with a good microphone tracks near the top of the range; multi-speaker panels, background music, and outdoor field recordings degrade in predictable ways. Numbers below are aggregated from the Whisper paper (Radford et al., OpenAI 2022), independent benchmarks (Open ASR Leaderboard), and VexaScribe's internal test corpus of known-transcript videos.

Source: Whisper paper (arxiv.org/abs/2212.04356) + internal testing; verified July 2026.

Practical read: for a talking-head vlog, a screen-recording tutorial, or a two-person podcast video, the SRT you get is broadcast-adjacent quality with light editing (a few minutes per hour to fix product names and proper nouns). For a panel discussion, an event-floor interview with crowd noise, or a field recording — expect meaningful cleanup time. Turn on speaker detection early; correcting speaker attribution after the fact is the most time-consuming edit.

Platform Requirements Matrix

Once you have the SRT, the last question is what your target platform actually accepts. Every major surface takes SRT, but a few (Instagram Reels most notably) do not accept soft-subs at all — they require burn-in. Others (TikTok) technically accept SRT via desktop uploader but reward burn-in with better feed engagement. Zoom and Wistia both auto-generate SRT internally but let you upload a higher-quality one from an outside source.

Sources: YouTube, Vimeo, LinkedIn, Facebook, TikTok, Instagram, Zoom, Wistia platform documentation; verified July 2026.

For a WCAG 2.1 SC 1.2.2 compliance workflow (Level A, prerecorded, June 2018) on YouTube, LinkedIn, or Vimeo, the correct path is soft-sub upload — a properly reviewed SRT beats auto-captions on the accessibility scoring rubric and lets viewers control their experience. For TikTok, Instagram Reels, and short-form social where captions drive silent-autoplay engagement, burn-in is the practical default. If you cross-post the same content to both surfaces, generate the SRT once (Job A), upload as soft-sub to long-form platforms, then burn-in with FFmpeg for the short-form cuts.

Sources & Verification

VexaScribe vs Alternatives

Common Use Cases

The six scenarios below cover the vast majority of what people mean when they search “video to SRT.” Match your situation to one and the workflow becomes concrete.

Related Guides

Convert video to SRT with VexaScribe (30 min free at signup, no card required) →